The Pliocene flora of Frankfurt am Main, Germany: taxonomy, palaeoenvironments and biogeographic affinities

The Pliocene flora of Frankfurt am Main described by Karl Mädler during the first half of the twentieth century is a key flora for the European Pliocene. In the present study, we revised the leaf fossil taxa described by Mädler and investigated plant material collected after Mädler’s publication. The revised and augmented floral list comprises seven new species and some new combinations of taxa described by Mädler. In total, 16 gymnosperm species in 15 genera and 73 angiosperm species (of which 15 could not be assigned to a genus) in 40 genera are recognised in the leaf record. Main characteristics of the flora are the high diversity of conifers, the diverse assemblage of exclusively deciduous Fagaceae, including six species of oaks, and the high diversity of Rosaceae. These features indicate cool temperate climatic conditions (comparable to Lugano in southern Switzerland). Angiosperm genera that are today confined to North America and/or East Asia (Eucommia, Magnolia and Sassafras) also are deciduous, whereas evergreen taxa are shrubs typical of the understorey (Buxus, Ilex, Pachysandra, Prunus lusitanica type) and Viscum. Eighteen taxa recorded in the Pliocene of Frankfurt am Main are today absent from western Eurasia and eastern North America, and 25 taxa are absent from western North America. This shows (i) a strong biogeographic link of the Pliocene flora of Frankfurt am Main with East Asia, (ii) surprisingly high levels of speciation (Pliocene endemisms) and (iii) that the European flora was more diverse in woody species shortly before the onset of major Pleistocene glaciations than today.


Introduction
Global cooling after the warm and mild phases of the Miocene (Zachos et al. 2001) led to a modernisation of north temperate floras (Mai 1995). The early part of the Pliocene (5.3-3.6 Ma) was characterised by warm conditions (ca. 3°C higher global surface temperatures) and higher sea levels (10-20 m; Ravelo et al. 2004) and slightly higher CO 2 concentrations (Beerling and Royer 2011). During the second part of the Pliocene, gradual cooling culminated in a significant intensification of northern hemispheric glaciation at ca. 2.75 Ma (Ravelo et al. 2004). Despite this, many exotic taxa persisted as relicts from older epochs (Mai 1995;Martinetto 2001), and modern diversity patterns of woody species (trees and shrubs) across the Northern Hemisphere were established only during and after the major Pleistocene glaciations (Magri et al. 2017). Today, species of the north temperate forest flora are distributed among western Eurasia, East Asia, western North America and eastern North America, approximately in the ratio 2:12:1:4 (Latham and Ricklefs 1993). These figures illustrate the strongly impoverished postglacial flora of Europe.
The Pliocene flora of Frankfurt am Main, Hessen, Germany (hereafter Frankfurt/M., Fig. 1), is a key flora for the European Pliocene. This flora has been repeatedly described (e.g. Geyler Electronic supplementary material The online version of this article (https://doi.org/10.1007/s12549-019-00391-6) contains supplementary material, which is available to authorized users. and Kinkelin 1887; Engelhardt and Kinkelin 1908;Mädler 1939) and is one of the richest macrofloras in the European Pliocene (Mai 1995). After Mädler's monograph, the Frankfurt BKlärbecken Flora^has never been revised.
After a previous revision of the Pliocene leaf flora of Auenheim, France , the main objectives of the present study were to re-assess previously published leaf morphotypes and to describe unpublished plant material of the Pliocene BKlärbecken Flora^site in Niederrad, Frankfurt/M. A major problem for the revision of this flora was that most of the material gathered since 1885 and housed at the Senckenberg Research Institute and Natural History Museum Frankfurt as a unique Bglass herbarium^ (Kräusel 1940) was destroyed in 1945 during World War II (Schaarschmidt 1980). Therefore, our objectives were (i) to investigate duplicate material, which could serve as a basis for the re-assessment of the Pliocene flora of Frankfurt (Mädler 1939), and (ii) to describe additional unpublished material recovered after World War II (Schaarschmidt 1980). Duplicates and unpublished material are housed in the collections of the Senckenberg Research Institute and Natural History Museum Frankfurt, Swedish Museum of Natural History, Stockholm and British Museum of Natural History, London Natural Science museum. Only leaf fossils are dealt with in this review, while the wood and palaeocarpological records (Supplementary material File 1) are left aside for future studies. The revised plant material was then used to assess biogeographic relationships of the flora of Frankfurt and to evaluate speciation events in the Pliocene. Finally, some conclusions regarding palaeoenvironments and palaeoclimate are presented.

Material and methods
Duplicate material and unpublished material of the Pliocene flora of Frankfurt When studying the collections at the Senckenberg Research Institute and Natural History Museum Frankfurt, one of us (Z.K.) noticed a sheet of paper with a list of material sent to the BNaturhistorisches Reichsmuseum Stockholm^, written by an anonymous keeper and without an author and date with a mechanical typewriter and containing out-of-date taxonomy (e.g. Planera ungeri). This short text lists Ginkgo adiantoides (leaf fragments), Abies pectinata (one seed cone), Picea latisquamosa (seed cone), Pinus montana fossilis (one seed cone), Pinus timleri (seed cone scale), Acer sp. (fruits), Betula (fruit bracts), Buxus sempervirens (leaves and fruit cupules), Carpinus grandis (leaf, involucre fragment), Carya alba fossilis (fruits), Carya ovata (fruits), Corylus avellana (fruits), Fagus pliocaenica (leaves), Ilex aquifolium (leaf remains), Juglans cinerea fossilis (fruits), Liquidambar pliocaenicum (infructescences), Planera ungeri (leaves) and Quercus (one leaf fragment). Additional notes on this sheet are written by hand and dispatched on the 19th February 1939 with a note that particularly conifer material would be welcome. Additions include Torreya nucifera fossilis (one fragmentary needle and epidermal preparation), Podocarpus kinkelinii (three leaves in one preparation), Abies pectinata fossilis (needle), Abies sclereidea (two needle fragments), Sequoia langsdorfii (three preparations), Taxodium distichum fossilis (one preparation), Libocedrus pliozänica (one twig), Carya globosa (one stone), Viscophyllum pliozänicum (leaf and epidermal preparation), Styrax obovatum (two stones), Stuartia europaea (one fruitlet) and Buxus sempervirens fossilis (one epidermal preparation). Further, material of Pinus timleri (two seeds), Viscophyllum miquelii (one leaf and one epidermal preparation) and Ilex aquifolium fossilis (one leaf and an epidermal preparation) was also sent to Stockholm. Hence, for the present re-assessment, we also concentrated on the material stored at the Stockholm museum (see Supplementary material File 1 for carpological material housed at the Stockholm museum).
Duplicates of the original or additional specimens from the Pliocene flora of Frankfurt/M. are likely to be found in more institutions. We have seen additional material at the British Museum of Natural History, London, distributed by Richard Kräusel. Such glass preparations and slabs with macrofossils are the only remnants of the once very rich collections of the Pliocene plant assemblage of Frankfurt that had been assembled since 1884 and were mainly known as Bglass herbarium ( Kräusel 1939).
Five papers dealing with material from the Pliocene of Frankfurt, namely on leaves of Fagus (Kvaček and Walther 1991;Denk 2004), seed cones of Pinus timleri (Kvaček et al. 2014a, b), Pseudotsuga loehrii (Kunzmann 2014), and leaves of Ginkgo (Denk and Velitzelos 2002) have been published after the classical paper by Mädler (1939); they are considered in the systematic part.

Preparation of the fossil material
The preparations of leaf fossils that formed the basis of the previous studies of the Pliocene flora of Frankfurt/M. followed a protocol described in Mädler (1939, p. 6). The sandy layers included Bleaf beds^, which allowed, after soaking in water and KOH, a mechanical separation of leaf fossils. The first extensive work of preparing slide preparations of fossil leaves was done by F. Kinkelin (Mädler 1939). The mummified leaf fossils were embedded between glass frames of the same size in glycerol and sealed with an unknown material. This procedure turned out to be less successful because glycerol evaporated through the framing. Later, Mädler attempted to improve the collection by spreading leaf fossils on glass, embedded in glycerol jelly and covered by a smaller plastic sheet, then sealed with Canada balsam (Mädler 1939, pp. 5-6).
Unfortunately, most of these glass preparations were destroyed as mentioned above. The procedure developed by Mädler was also employed for preparations of the newly excavated material obtained by Richard Kräusel who transferred it for further preparation work to Wolfgang Haas in 1956 (see personal letter by Kräusel to Haas in the archive of the Department of Palaeobotany of the Senckenberg Museum; Volker Wilde, personal communication 2014). These specimens are quite well preserved and used in the present study. Only in rare cases the embedding material (glycerol jelly) partly dried out.
The collection of new preparations is now housed in the Section of Palaeobotany, Senckenberg Research Institute and Natural History Museum Frankfurt. It is arranged systematically according to the fossil species. The inventory numbers (SF.B numbers) follow this systematic arrangement. The numbers were scratched onto the glass slides by an anonymous keeper.
For the present study, only a small part of the glass preparations of macrofossils has been re-opened to get leaf lamina fragments for cuticle preparations. Mädler (1939) used Schulze solution in combination with ammonia for maceration and a similar method was employed for the present study except for using 5% KOH instead of ammonia (see protocol in Kvaček et al. 2008). The preparation of cuticles was not always successful, namely in case of delicate cuticles of deciduous foliage when highly diluted Schulze reagent was used for maceration. Although not attempted here, fluorescence microscopy may turn out to be promising in future research.
The newly revised specimens and epidermal slides are housed in the collection of the Senckenberg Research Institute and Natural History Museum Frankfurt (numbers with prefix SF.B). Some specimens of the original material available for this study are housed at the Swedish Museum of Natural History, Stockholm (numbers starting with S) and the British Museum of Natural History, London (specimens numbered with prefix V).

Inferring palaeoclimate and palaeoenvironments
Palaeoclimate and palaeoenvironment estimates were made using Climate Leaf Analysis Multivariate Program (CLAMP) and Integrated Plant Record (IPR) vegetation analysis. The CLAMP method (e.g. Wolfe and Spicer 1999) uses physiognomic characteristics of the studied plant assemblage of Frankfurt/Main presented in Supplementary material File 2 and the physiognomic and gridded meteorological calibration datasets from 144 sites BPhysg3br/GRIDMet3br^(see Spicer et al. 2009) selected by the statistical tool published by Teodoridis et al. (2012). The CLAMP analysis was performed using a web application designed by Yang et al. (2011), which is free to access at the official CLAMP website (Spicer 2011(Spicer -2019 for details, see Supplementary material File 2).
The IPR-vegetation analysis is a semi-quantitative method (Kovar-Eder et al. 2008) to evaluate zonal vegetation. This method scores key component groups (functional types) of the integrated plant fossil record. This technique has been verified using modern plant assemblages from SE China and Japan ) and has recently been tested by the second author using living European and Caucasian plant assemblages (Bohn et al. 2004).

Systematic palaeobotany
The leaf taxa listed below were partly described by Karl Mädler in his original study (Mädler 1939) and were dealt with by various authors in subsequent studies (mentioned in the synonymies). Several novelties for the Pliocene flora of Frankfurt/M. have been provisionally indicated in the collections (possibly by R. Kräusel); other undescribed taxa are here suggested for the first time. Only a few separate papers have appeared recently, which are related to the Pliocene flora of Frankfurt/M. (e.g. Kvaček and Walther 1991;Denk 2004), partly based on the material from other sites (e.g. Pinus timleri by Kvaček et al. 2014a, b; Ilex geissertii by Kvaček et al. 2008). We prefer to assign the fossils to fossil-species instead of using names of modern taxa followed by Bfossilis^as Mädler (1939) did. This usage is still maintained in current papers dealing with Pliocene taxa (e.g. Geissert 1973; Mai and Walther 1988;Hably and Kvaček 1997;Knobloch 1998). In a few cases, when leaf morphology and cuticle structure are virtually identical with modern species, e.g. Acer platanoides, we called the fossiltaxon Acer aff. platanoides, indicating that the fossil most likely represents the lineage leading to the modern species, but perhaps not the species itself.
The arrangement of taxa follows the system based on the classification suggested by the Angiosperm Phylogeny Group (APG IV 2016). The collections studied include also fossils of reproductive organs (seed cones, winged carpological material), which are briefly mentioned in the systematic section but not treated in detail. Only selected foliage specimens of the entire collection are listed and illustrated, mainly those studied by epidermal anatomy. Descriptions are only provided for taxa that have not been described by Mädler (1939).
Remarks: No additional information to that published by Mädler (1939) has been obtained by the study of the newly collected material. The description corresponds to the previously published data in Mädler (1939). For comments on the taxonomy, see Kvaček et al. (2008). The material from the new collection was used for a study of stomatal density as a proxy for atmospheric CO 2 concentration (Retallack 2001). Denk and Velitzelos (2002) noted that the abaxial epidermis of Ginkgo from Frankfurt shows only faint papillae, while in late Miocene material from northern Greece, the papillae were prominent and overarching the stomata.
Remarks: Leafy shoots embedded in glass preparations correspond in morphological features to the type material from the Bohemian Miocene (see Kvaček 1976). Macerations of needles have not been successful, and the obtained preparations reveal only sub-macerated abaxial stomata orientated perpendicularly or obliquely to the needle length. Mädler (1939) Mädler (1939). These records are not treated in the systematic part below. For the seed cones of Pinus, the revisions by Mai (1986) and Kvaček et al. (2014a), b) and for Keteleeria loehri (Geyler et Kinkelin) Kinkelin, the revision by Kunzmann (2014) are accepted.
Picea A. Dietrich Picea omoricoides C.A. Weber Fig. 15d-f 1898 Picea omoricoides C.A. Weber,p. 5, Material: One slide V. 17164 identified as Picea, cuticle preparations S117938-117942 coll. Stockholm. Description: Macerated needles up to 2 mm wide, incomplete in length, rounded at apex, one side without stomata, ordinary cells coarsely undulate, ca. 15 μm wide, hypodermal cells straight-walled, similar in size, opposite side with longitudinally arranged narrow epidermal cells with deeply and regularly undulate anticlinal walls, stomata in five to eight rows, longitudinally aligned, bicyclic, subsidiary cells four to eight, with shallow undulate to pitted anticlinal walls, polar quadrangular, usually bordering two adjacent stomata, lateral subsidiary cells elongate, on one side usually shared among two adjacent stomata. Remarks: The epidermal topography corresponds to that described for the Pliocene records of Picea omoricoides from Germany (Gerstungen and other sites, Mai and Walther 1988) and North Bohemia (Cheb Basin, Vildštejn Formation, Bůžek et al. 1985;Teodoridis et al. 2017).
Pinus Linnaeus Pinus sp. Fig. 2m 1908 Stiel von Acer, Engelhardt and Kinkelin, p. 296, pl. 34, fig. 11b 1939 Kurztriebe von Pinus;Mädler, p. 35, pl. 1, fig. 36 Material: SF.B 11433 and 11436 (two 2-needled fascicles), two missing specimens of 2-and 3-needled fascicles quoted by Mädler (1939). Description: See Mädler (1939, p. 35). Remarks: Needles in double fascicles have not been anatomically studied being enclosed in glycerol jelly between glass and plastic plates. They may belong to some of the associated seed cones but a straight-forward connection cannot be established.  Mädler (1939) assigned these peculiar, long needles to Podocarpus L'Héritier ex. Persoon and mentioned great difficulties in investigating the epidermal parts. Also the presently studied material (SF.B 11419) was not easy for preparation of cuticles. The lower epidermis remained firmly attached to the leaf tissue in spite of strong maceration. Cathaya is fairly similar to the fossils discussed but lacks papillae within stomatal bands and has quadrangular polar subsidiary cells (Kunzmann 2014). Delicate deciduous needle foliage of Pseudolarix Gordon may recall Pseudotsuga kinkelinii but has stomatal zones smooth. A long needle from the Pliocene of Auenheim corresponding also in the epidermal structure to the Frankfurt/M. specimens was misinterpreted as Cathaya (see Kvaček et al. 2008, p. 8, pl. 1, fig. 10, pl. 16, fig. 6, as Cathaya sp.). Szafer (1961), p. 26, pl. 7, figs. 1-6) referred needle fragments from the late Miocene of Poland with similar epidermal structure to Podocarpus. Foliage of Pseudotsuga kinkelinii co-occurs with seed cones recently investigated by Kunzmann (2014) and assigned to Pseudotsuga loehrii (Geyler et Kinkelin) L. Kunzmann.

Pseudotsuga
By its acute leaf apex Pseudotsuga kinkelinii is similar to the extant North American species. However, most modern species have shorter needles. Pseudotsuga macrocarpa (Vasey) Mayr from southern California is most similar by seed cone morphology to P. loehrii (Kunzmann 2014) as well as to P. kinkelinii by needles up to 50 mm or even 60-80 mm long with acute tips. This modern species grows also in riparian habitats along with Acer macrophyllum Pursh and Populus trichocarpa Torrey et A. Gray (Barbour 1988 fig. 19, pl. 9, figs. 9-12^. Florin probably intended this material to be included in a more extensive revision on fossil Cephalotaxus and Taxus (see Seward and Edwards in Boulter and Kvaček 1989, p. 62), but this revision remained unpublished. According to its epidermal structure, the specimen differs from T. inopinata Givulescu (1973Givulescu ( , 1975 from the Romanian Pannonian and from Taxus sp. 2 sensu Kvaček (1984) from the Pliocene of North Bohemia because the abaxial papillate zones along the stomatal bands are wider, reaching halfway to the stomatal bands. In the Romanian material, papillae are confined to the stomatal areas and also in the fragment from the Bohemian Pliocene/ Pleistocene described as Taxus sp. 2 sensu Kvaček 1984), papillae densely covering the stomatal bands are almost lacking on adjacent non-stomatal zones (Kvaček 1984, fig. 4a, e). The Pliocene record of Taxus from Willershausen, identified as Taxus baccata L. fossilis (Straus 1952, p. 20, pl. 4, figs. 6-7, pl. 7, figs. 28, 30) does not differ from that of Frankfurt according to the epidermal anatomy and may belong to the same fossil species. We follow Macovei (2013), who recommended not using independent fossil species names for remains that represent close ancestors of Taxus baccata L. The Oligocene T. engelhardtii Kvaček differs by the fully papillate abaxial medial area (Kvaček 1976(Kvaček , 1984 Material: The type material is destroyed and no new topotype material is available at SF. Remarks: Although no material of this conifer described by Mädler (1939) has been recovered in the collections available, the description and illustrations are sufficient for the recognition of this fossil species in the Pliocene flora of Frankfurt/M. Similar needles were recorded in the Romanian Pannonian and assigned to Cephalotaxus pliocaenica by Givulescu (in Givulescu and Olos 1973, p. 32, pl. 12, figs. 4-6, pl. 14, pl. 1, pl. 15, figs. 7-8). A single fragment from the early Miocene Cypris formation of north Bohemia (Kvaček 1984) shows sclereids in the leaf tissue not noticed in the type material of Cephalotaxus pliocaenica Mädler, although it matches the illustrations and descriptions of Mädler (1939)  Material: All original material listed by Engelhardt and Kinkelin (1908) and Mädler (1939) was destroyed except two needle and epidermis preparations kept at NRM (S082693, S082791). Another specimen was found at the BMNH, V. 17176. No new material is available at SF. Description: See Mädler (1939, pp. 13-14).
Remarks: The fossil leaf material corresponds in most respects to extant T. nucifera distributed in Japan. A more detailed comparative study of foliage between living and fossil representatives of Torreya is needed to establish their relationships. It is probable that dispersed needles and seeds in the Pliocene of Frankfurt/M. represent a single biological species according to the whole plant concept of Kvaček (2008). In any case, the foliage deserves an independent fossil species name, as done for the seeds (T. schulzii Gregor, van der Burgh, Peters, Pingen- Gregor et al. 2000).

Angiospermae
Magnoliaceae Jussieu  (Kräusel and Weyland 1959, as Papilionaceophyllum liblarense) and most other Miocene occurrences in Europe (see Kovar-Eder and Hably 2006;Schneider 2004, as Falcicutis varians Schneider), the fragment recovered here as well as the illustrations by Mädler (1939) from the same site differs in glabrous laminas. Thus, the fossil species M. liblarensis may turn out to be heterogeneous or is highly variable in pubescence, as documented by Fischer and Butzmann (2000). Several living Magnoliaceae with similar epidermal patterns (undulate anticlinal walls, rare or lacking pubescence) have been noticed in the modern cuticle collection of Z.
Kvaček   Description: Leaves long petiolate, petiole > 15 mm long, lamina ovate, 80 mm long, > 40 mm wide, entire-margined, apex bluntly acute, base widely cuneate, venation camptodromous, primary vein stout at base, much thinner apically, secondary veins irregularly spaced, more closely spaced at base than towards apex, bent, forked and looping at margin, intersecondaries rare and single, intercostal tertiary veins straight, perpendicular to secondary veins. Adaxial cuticle smooth, hairless, polygonal cells with straight anticlines without thickenings 20-30 μm in diameter, abaxial cuticle very thin, ordinary cells domed, stomata not sunken, brachyparacytic, guard cell pairs broadly elliptic to transversally elliptic, 15 μm long, stomatal ledges slightly thickened in the middle, forming spindle-shaped narrow outer aperture. Mesophyll tissue filled with lens-shaped oil cells 25-50 μm in diameter. Remarks: Thin cuticles, mesophyll oil cells and the preserved epidermal structure refer these leaf fragments to the Magnoliales. The overall venation and the morphology compare well with foliage of deciduous magnolias. The fragment of M. liblarensis described above clearly differs in epidermal patterns, namely the undulate anticlines, non-papillate surface and distinct roundish stomatal complexes. Magnolia waltheri resembles by its stomata patterns Laurophyllum kinkelinii Kvaček (2004)  Magnolia waltheri matches best in epidermal anatomy (thin cuticles, straight-walled anticlines) and morphology (widely cuneate leaf base, tertiary venation) modern deciduous magnolias, such as Magnolia kobus DC. (Japan). Magnolia salicifolia (Siebold et Zuccarini) Maximowicz (Japan) resembles the fossil species also by its doomed abaxial cells. Magnolia waltheri probably represents foliage belonging to seeds assigned to M. cor Ludwig and described from the Pliocene flora of Frankfurt/M. by Mädler (1939). According to Mai (1975, p. 564 Description: Leaves petiolate, petiole up to 19 mm long, lamina ovate, 80 mm long, 23-40 mm wide, entire-margined, apex bluntly acute, base widely cuneate, venation campto-dromous, primary vein stout at base, much thinner apically, secondary veins irregularly spaced, more closely spaced at base than towards apex, bent, forked and looping at margin, intersecondaries rare and single, intercostal tertiary veins straight, forming acute angles with secondary veins. Adaxial cuticle smooth, hairless, reflecting polygonal cells with straight anticlines without thickenings 20-30 μm in diameter, abaxial cuticle very thin, ordinary cells doomed, stomata not sunken, brachyparacytic, guard cell pairs broadly elliptic to transversally elliptic, 15 μm long, stomatal ledges slightly thickened in the middle, forming spindleshaped narrow outer aperture. Mesophyll tissue filled with lensshaped oil cells 25-40 μm in diameter. Remarks: Similar thinly cutinised leaf compressions with oil cells may belong to the Lauraceae, most probably to deciduous plants. The preserved compressions show similar morphological features as the leaf fragment assigned to the same taxon from the Pliocene of Auenheim ).

Smilacaceae Ventenat
Smilax Linnaeus Smilax sp. Fig. 3d Material: SF.B 11528, 11529. Description: Leaves incomplete, petiolate, lamina broadly ovate, base shallowly cordate-concavo-convex, apex missing, margin entire, venation campylodromous, primary veins 5, secondary veins very thin forming wide meshes with higherorder vein matrix ascending towards the margin and looping far from it. Remarks: The few foliage remains at hand surely belong to the Smilacaceae because of the characteristic gross morphology and venation. Both specimens fall within the morphological variation of Smilax sagittifera Heer sensu Hantke (1954) and S. weberi Wessel. Similar leaf shapes are found in a great number of modern species (Denk et al. 2015). Description: Narrowed strip-like leaf fragments, entiremargined and parallel veined. Remarks: Grass-like foliage fragments not identifiable to a genus noted also by Mädler (1939, p. 49 Adaxial cuticle thick, ordinary cells 38 μm in diameter, polygonal, anticlinal cell walls smooth, straight to little curved, abaxial cuticle thick, ordinary cells the same as in adaxial cuticle, stomata broadly elliptic, 50 × 40 μm in size, irregularly orientated, anomocytic (to cyclocytic), with a thick stomatal ring inside, I-pieces at poles, aperture between guard cells linear, short massively cutinised trichomes simple, subulate, 50-80 μm long and 30 μm thick, on margin. Oil cells in the mesophyll tissue disc-shaped, 30-50 μm in diameter.
Remarks: The epidermal structure including the stomata corresponds to the modern representatives of Pachysandra (see also Baranova 1980). Fossils of Pachysandra have been reported from the European Palaeogene as rare seeds (e.g. Mai and Walther 1985, as P. ascidiiformis Mai) and pollen (Krutzsch 1966; also noted in the Eocene of Axel Heiberg Island, McIntyre 1991). To our knowledge, fossil foliage of Pachysandra has not previously been known. Straus (1992, p. 58) mentioned a record of Pachysandra for the Pliocene Willershausen flora without further description or documentation. Both morphological and leaf anatomical traits of the single compression available indicate its affinity with this evergreen to semi-evergreen subshrub distributed by two or three extant species (P. terminalis Siebold et Zuccarini, P. axillaris Franchet and P. stylosa Dunn) in East Asia and by one species (P. procumbens A. Michaux) in southeastern USA. The fossil at hand is best comparable by its glabrous lamina and the leaf form with P. terminalis distributed in Japan and China (Gansu, Hubei, Shaanxi, Sichuan, Zhejiang) in shady and damp land in forests at altitudes between 1000 and 2600 m a.s.l. (Wu and Raven 2008, p. 331). Two other extant species from East Asia; P. axillaris and P. stylosa differ by the rounded to cordate leaf base and thick pubescence, P. procumbens from the USA by the coarsely toothed margin and hairy leaves, both adaxially and abaxially.

Altingiaceae Lindley
Liquidambar Linnaeus Liquidambar europaea A. Braun Fig. 3j 1836 Liquidambar europaea A. Braun in Buckland,p. 513 Material: SF.B 12158. Description: Leaf incomplete, pentalobate, lobes with glandular crenulate margin. Remarks: The single specimen of Liquidambar europaea in the Pliocene of Frankfurt/M. confirms the rarity of this element in this flora. It is also known from the carpological record of Frankfurt (Supplementary material File 1). Liquidambar is lacking in the Pliocene of Auenheim  and Willershausen (Knobloch 1998). The pentalobate leaf form of this fossil species is more common than the trilobate one in the late Neogene of Europe (see Knobloch and Kvaček 1976;Hummel 1983). Zhilin (1974) noted that according to Smith (1967) Liquidambar styraciflua produces two leaf forms on every branch differing in the depth of the lobes and the length of the petioles. This variation does not concern the number of lobes, though. According to Hummel (1983) Remarks: A single fragmented leaf compression belongs to this common deciduous element of the northern hemispheric Cenozoic flora based on its morphology (venation palmate, 5veined, margin regularly closely uniformly crenate). Fossil populations slightly differ in fruit morphology as shown in foliage shoots with attached fruits (Smiley and Rember 1985;Kvaček and Konzalová 1996). It is closely similar to the modern species C. japonicum Siebold et Zuccarini from Japan and China, a deciduous tree inhabiting riparian forests. Derivatio nominis: Referring to the geological age of the Klärbeckenflora, Frankfurt/M. Description: Leaflets detached, sessile or exceptionally with very short petiolule, 22-40 mm long, 12-33 mm wide, narrowly to broadly ovate, base asymmetrical, rounded to cuneate, apex broadly acute, margin irregularly, simple bluntly serrate except entire-margined base, teeth simple, shallow asymmetrically rounded, venation semicraspedodromous, craspedodromous, primary vein straight or slightly curved, secondary veins in up to seven pairs with single intersecondaries, looping widely along the margin, lateral veinlets entering the margin into teeth, tertiary veins forming asymmetrical meshes with higher-order vein matrix. Epidermal structure of the holotype: adaxial cuticle smooth, ordinary cells polygonal, 25-40 μm in diameter, anticlinal walls slightly bent, simple thin trichomes up to 150 μm dispersed on veinlets, abaxial cuticle smooth, ordinary cells with curved or bent anticlinal walls, mostly 30 μm in diameter, over veins quadrangular, 12-15 μm wide, 25-75 μm long, solitary elliptical thin-walled trichome bases 25 × 15 μm in size, stomata irregularly disposed, elliptic, irregularly sized, average 25 μm long, 15 μm wide, ? anomocytic, stomatal ledges slightly thickened, aperture boat-shaped, slit linear. Remarks: Similar leaflets were described as Gleditsia by Guo and Zhou (1992) from the Miocene of China and by Kvaček et al. (2011) from the Miocene of southwestern Europe (Arjuzanx). Gleditsia suevica Rüffle (1963) and the foliage described as Podogonium oehningense (Koenig) Kirchheimer sensu Rüffle (1963) of the Miocene of Germany show a similar structure of the adaxial as well as abaxial epidermis and pubescence but differ by the entire margin and petiolulate leaflets. The modern Gleditsia caspica Desfontaines produces similar foliage and might be the closest extant relative of the fossil taxon; it is a typical element of the lowland Hyrcanian forests associated with other relic elements, such as Pterocarya, Albizzia and Parrotia (Denk 1998). Gleditsia pliocaenica differs from G. caspica in broader asymmetrically ovate leaflets and a distinctly crenulate lamina.
aff. Podocarpium A. Braun aff. Podocarpium sp. Fig. 4c Material: SF.B 12432. Description: Leaflet, sessile, narrow ovate, 25 mm long, 9 mm wide, rounded at base and apex, margin almost entire, venation brochidodromous, primary vein slightly bent, secondary veins steep, in six pairs, with rare single intersecondaries, tertiary veins reticulate. Remarks: The single specimen is similar to Gleditsia pliocaenica based on its venation but might belong to a different genus of the legumes. Leaflets of Podocarpium podocarpum (A. Braun) Herendeen widely distributed in the European Neogene differ in fully entire margins and one stronger basal, asymmetrically disposed, secondary vein (Bůžek 1971, p. 98 1971). In the early Miocene flora of northern Bohemia, this foliage is accompanied by long-stalked samaras with a calyx remain above the stalk base (Bůžek 1971 Kvaček and Walther (1991) and Denk (2004).
The cuticle structure of specimens was not accessible.
Derivatio nominis: The species epithet refers to the resemblance of the leaf with Viburnum opulus L. Description: Leaf simple, petiolate, petiole thin, lamina chartaceous, shallowly trilobate, apex broadly acute, margin shallowly bluntly simple dentate-lobed, venation basal actinodromous, midrib thin, only slightly thicker than secondaries, basal veins departing at an angle of 40°, higher secondary veins subparallel and widely irregularly spaced, simple, looping and curved along margin, tertiary veins irregularly reticulate. Adaxial and abaxial cuticles thin, adaxial cuticle faintly striate, unspecialised cells irregularly polygonal, 25-50 μm in diameter, with curved anticlinal walls, abaxial cuticle densely finely papillate, papillae granular, ca. 8 μm in diameter, papillae larger towards leaf margin, densely distributed on stomatal areas, forming rings around hardly visible elliptic stomata, stomata 15 μm long and 10 μm wide. Remarks: The leaf compression corresponds in leaf morphology to Acer and Viburnum. Acer obtusifolium Sibthorp et Smith (eastern Turkey, Syria, and Lebanon), Acer sect. Acer is closely similar to the fossil species in leaf epidermal characteristics (Grimm et al. 2007) and in overall leaf shape but differs from the fossil species in its sub-entire margin. The previously applied species name Acer brachyphyllum Heer (Mädler 1939, p. 114, pl. 9, figs. 9-10) is reserved for specimens that belong morphologically to A. tricuspidatum Bronn f. brachyphyllum (Heer) Procházka et Bůžek (1975, p. 28), although epidermal characters of the type specimens from the Sarmatian of Öhningen are not available. Viburnum opulus var. opulus and a closely related American species V. trilobum Marshall, often considered as V. opulus var. americanum Aiton imitate Acer in leaf morphology but clearly differ in epidermal anatomy (non-papillate, striate abaxial cuticle, a different kind of anomocytic stomata). Remarks: The material shows affinities to the North American members of sect. Pavia (Mill.) Persoon based on the long petiolulate leaflets and epidermal anatomy. The leaves were probably produced by the same species as the fruit and seed fragments described from the same site by Engelhardt (in Engelhardt and Kinkelin 1908). Similar leaflets were described from the Pliocene of Willershausen (Straus 1930, as Aesculus cf. pavia Linnaeus, Knobloch 1998, as A. velitzelosii Erw. Knobloch). Some other records of Aesculus foliage not preserved with leaf anatomy (Iljinskaja 1968 B 11664, 11665, 11666, 11667, 11668. Description: Leaflets shortly petiolulate, petiolule max. 5 mm long, lamina entire-margined, asymmetrically ovate to oblong, up to > 100 mm long, 24-32 (−45) mm wide, apex acuminate, base truncate, rounded to widely cuneate, venation brochidodromous, primary vein thick, slightly curved, secondary veins in more than 13 pairs, secondary veins alternate to sub-opposite, slightly s-shaped, more closely spaced towards apex, slightly curved, looping near margin, single intersecondary veins commonly present, tertiary veins alternate percurrent, quaternary veins regular reticulate. Epidermal anatomy (sample 11668), adaxial cuticle faintly striate, reflecting polygonal outlines of straight-walled ordinary cells 20-25 μm in diameter, anticlinal walls straight, solitary trichome bases simple, rounded, 12 μm in diameter, abaxial cuticle irregularly finely striate, rarely with preserved rounded to elliptic head up to 120 μm long and 70 μm wide, cell outlines not observable, stomata elliptic, anomocytic, all of the same size, 15-20 × 12 μm, stomatal ledges slightly thickened, outer cavity elliptic, boat-shaped, 12-15 μm long, rare simple filamentous trichomes up to 120 μm. Remarks: This fossil species has been commonly recorded in the European Neogene and is quite characteristic in leaf morphology (e.g. Bůžek 1971). The epidermal structure clearly differs by peltate glandular trichomes from the modern species of Juglans including J. regia Linnaeus, usually cited as a modern analogue. A typical, broader form of BJuglans^acuminata from Arjuzanx (Kvaček et al. 2011) showed also peltate glandular trichomes in an unmacerated specimen embedded between glass slides. Morphologically similar but narrower impressions from the Pliocene of Willershausen (Knobloch 1998) were assigned to Cedrela heliconia (Unger) Erw. Knobloch, those from the Sarmatian from Bulgaria to Cedrela attica (Unger) Palamarev et Petkova (Palamarev et al. 2005). The abundant foliage of J. acuminata is associated with seeds of the Toona type at the early Miocene site Čermníky in North Bohemia (Bůžek 1971, p. 102, pl. 24, figs. 20-21, as Fructus vel semen). The affinity of the record from Frankfurt/M. to the Meliaceae is clearly demonstrated by the indumentum of the leaf. Foliage of various genera of Meliaceae is morphologically similar (e.g. Toona (Endlicher) M. Roemer of Asia and Australia, Cedrela P. Browne, Swietenia Jacquin of Mexico and Central America) but a lepidote indumentum is rarely present (e.g. in Aglaia Loureiro from the Philippines, Indonesia to Taiwan). Most modern species of Aglaia distributed in East and Southeast Asia (to Australia) have a lepidote foliage and match also in leaf morphology (e.g. A. lawii (Wight) C.J. Saldanha from Taiwan, Yunnan and elsewhere in southern Asia). Contrary to the mentioned evergreen representatives of Meliaceae, the studied fossil specimens differ by a papery (obviously deciduous) leaf lamina. In other occurrences of BJuglans^acuminata, most specimens are isolated leaflets and complete leaves are extremely rare (Heer 1859). We are unable to assign BJuglans^acuminata to a particular genus of the Meliaceae. Fossil Cedrela merrillii (Chaney) P. Brown from the North American Cenozoic differs by the wavy margin of the leaflets and associated seeds as documented by Meyer and Manchester (1997), 131, pl. 54, figs. 1-12 Kräusel 1939, p. 86, pl. 8, fig. 2 andin Kräusel 1940, text-fig. 9b).  (Fig. 14b).

Acer vitiforme
Description: Leaves elliptic, base widely cuneate and narrowing towards stout petiole, apex rounded, margin entire, venation basal acrodromous, five primary veins straight, venation of higher-order ramified. Remarks: The epidermal anatomy and relationship of the fossil species have been described and discussed several times (Knobloch and Kvaček 1976;Kovar-Eder and Krainer, 1991;Kvaček et al. 2008).
Remarks: For detailed description and comments, see Kvaček et al. (2009).
Angiospermae incertae familiae Dicotylophyllum Saporta The leaf remains described below cannot be assigned to particular families and genera in most cases. Remarks are only provided in cases where closer similarities to some taxa are observed. Cuticle preparations are available in rare cases, when leaf specimens were accessible to remove cuticle samples.

Discussion
Age of the flora of Frankfurt/M.
The Pliocene flora of Frankfurt/M. was discovered in 1884 and numerous plant fossils were collected after 1885, when Fig. 19 a Viscum miquelii (Geyler et Kinkelin) Czeczott, leaf cuticle, S a sandy clay lens had become exposed during the construction of the clearing basin of the sewage treatment plant in Frankfurt/Niederrad. The depositional setting of the flora makes a closer age determination difficult, and various authors have suggested a late Miocene, early and late Pliocene, or even Quaternary age for the plant assemblage of Frankfurt/ M (cf. Mädler 1939). The general regional lithology comprises sand and gravel on top of clays or surrounding clay lenses and lignites (Martini et al. 2011). Zagwijn (1960) established a palyno-stratigraphic framework for the late Miocene to early Pliocene in the Netherlands and made a clear distinction between the early Pliocene Brunssumian and the late Pliocene Reuverian. Previously, Reid and Reid (1915) Kemna and Westerhoff (2007) objected that the Sequoia-type pollen peak found by Zagwijn was not a reliable feature to identify the Brunssumian stage but acknowledged that the presence of Symplocos and Edmundipollis edmundi were more diagnostic of the first half of the Pliocene. According to Stuchlik et al. (2014), the pollen taxon Edmundipollis edmundi has closest botanical affinities with the extant genus Diplopanax (Cornales, Nyssaceae). Both Symplocos and Diplopanax are absent from the leaf and carpological record of Frankfurt (Supplementary material File 1). Mohr (1986) investigated a 2.5-m pollen profile from Willershausen und based on the rare occurrence of Symplocos, and the absence of other Bolder^elements such as Sapotaceae, Engelhardioideae suggested a Reuverian age for the flora of Willershausen. This age was accepted by Knobloch (1998). Mai (1995), based on the presence of warmth-loving relicts, distinguished a BFlorenkomplex Brunssum^and a BFlorenkomplex Reuver^. While placing Frankfurt/M and the Bmain fossiliferous horizon^of Auenheim (Alsace) in the Floren-komplex Brunssum, Mai (1995) considered the floras of Willershausen and Berga to be of Reuverian age.
Based on the work by Geissert (1972) and Geissert et al. (1990), Kvaček et al. (2008), fig. 2) correlated the Pliocene leaf flora of Auenheim (Alsace, France) with the BVillafranchian^strata of Geissert (1972) and suggested a late Pliocene age (Reuverian). The leaf flora of Frankfurt/M is virtually identical with the one from  Climate data from Thompson et al. (1999aThompson et al. ( , 1999bThompson et al. ( , 2000 and Fang et al. (2011) MAT = mean annual temperature in C°, MTCM= mean temperature of the coldest month in C°, EA = East Asia, NA = North America File 1). In general, there is a remarkable overlap between taxa found in the leaf and the carpological record (Table 3). Differences between the two records are mainly due to the lack of herbaceous species in the leaf fossil record. Of these, about 10 are strictly aquatic species, while most of the others can thrive in different habitats including wet meadows, riparian sites, forest undergrowth, etc. (Supplementary material File 1). One exception may be the genus Scleranthus, which is more common on dry sites and as pioneer in open places (Oberdorfer 1979). In addition, a number of woody taxa are absent from the leaf record as well. These include Engelhardioideae,  a Walter-climate diagram. Lugano in southern Switzerland has a typical temperate climate, Cf climate according to the Köppen-Geiger climate classification. There is no period during the year where rainfall is a limiting factor; summer months are characterised by highest temperatures and precipitation. b Walter-climate diagram. Santander in northern Spain has a sub-Mediterranean climate. Rainfall is limited during the summer months but there is no arid period during the summer as in Mediterranean climates. c Climate variables for Lugano indicating, among others, the mean temperature of the coldest month. All figures from Lieth et al. (1999) Torreya have highly relictual extant distributions either disjunct between North America and East Asia or confined to one of them (Table 1). The same is true for many angios p e r m t a x a s u c h a s S a s s a f r a s , L i q u i d a m b a r, Cercidiphyllum, Eucommia and Trichosanthes.
Besides, several relict areas in western Eurasia harbour many of the plants recorded from Frankfurt. For example, in the Euxinian-Hyrcanian region (northern Turkey, Transcaucasia, to northern Iran), a great number of western Eurasian relict taxa are found (Buxus, Celtis, Gleditsia, Ilex, Parrotia, Pterocarya, Zelkova). Remarkably, all the genera reported from Frankfurt are growing in the Botanical Garden of Batumi (Georgia; Sharadze 1987). Trees such as Cryptomeria, Cunninghamia and Eucommia further have naturalised in the hinterland of Batumi (T. Denk, personal observation). A few relict species are confined to small areas in the East Mediterranean region (Liquidambar orientalis). Kvaček et al. (2008) noted that most plants from the late Pliocene flora of Auenheim (Alsace) belong to mesophytic representatives of mixed coniferous and broad-leaved deciduous forests. The same is true for the Pliocene flora of Frankfurt. Overall, the Pliocene flora of Frankfurt can be viewed as an impoverished northern hemispheric Neogene forest flora with numerous exotic elements.
A survey of climatic parameters of conifer taxa that are today absent from western Eurasia but persisted in East Asia and/or North America shows that most of these genera tolerate fairly low mean temperatures of the coldest month (Table 2). A single monotypic genus, Glyptostrobus, is currently restricted to areas with coldest month mean temperatures close to the freezing point. However, observations and experiments on freezing tolerance of modern conifers suggest that the realised niches in relict conifers are reflecting only a very small part of their fundamental niches (Sakai 1971;Bannister and Neuner 2001). Assuming a modern distribution of rainfall and temperature over the year (rain maxima in summer, coldest in winter), the modern climate of Lugano in the south of Switzerland would provide conditions to accommodate the Pliocene flora of Frankfurt (Fig. 21 a, c). In contrast, several   Fig. 21b; San Sebastian, Batumi) have different rainfall and temperature patterns, with maximum rainfall in winter. In a previous study, Thiel et al. (2012) used leaf physiognomic characteristics of dicotyledonous angiosperms (CLAMP, Spicer 2011-2019 to infer the palaeoclimate of Frankfurt. The reconstructed MAT (12.2 ± 1.2°C), CMMT (2.3 ± 1.9°C) and MAP (979-1333 mm) closely match the conditions in sheltered areas south of the Alps (see above Lugano, Ljubljana, Udine;Lieth 1999). It is noteworthy that highly similar climate estimates are obtained from modern climate envelopes of North American and East Asian conifers ( Table 2) and from leaf physiognomic signal of broad-leaved angiosperm leaves.
Speciation and extinction in the late Pliocene of Europe Many cool temperate northern hemispheric tree genera that were abundant during the Pliocene went extinct in Europe or became restricted to small relict areas in western Eurasia during the Pleistocene (Svenning 2003;Magri et al. 2017). For example, Sciadopitys, Taxodioideae, Tsuga, Carya, Pterocarya and Zelkova persisted in northern Italy until less than one million years ago (Magri et al. 2017). The majority of taxa recorded for the Pliocene of Frankfurt persisted in western Eurasia until today (Table 1). According to Svenning (2003), increased cold and drought tolerance are important factors that determined whether tree taxa are still widespread in Europe, survived in relict areas or went extinct. Modern counterparts of extinct conifers in the Pliocene of Frankfurt are relatively cold tolerant (see above; Table 2) but are much more sensitive to cold than widespread modern cool temperate tree genera of Europe (cf. table 4 in Svenning 2003). Several new fossil species recognised in the present study and by Mädler (1939) are noteworthy, as they might reflect a Pliocene radiation that did not persist into modern times. Mädler (1939) established eight new species based on fruits and seeds (Table 3), while in the present study, seven new species were recognised (Table 4). Some of the taxa recognised as new species belong to genera that are still widespread in western Eurasia. For example, in the widespread northern hemispheric genus Acer, Acer dombeyopsis is morphologically closely similar to sections Pentaphylla and Lithocarpa, native to the Himalayas and East Asia. Acer viburnoides and vitiforme are extinct species in the Eurasian sect. Acer. Acer is still quite diverse in the flora of western Eurasia with ca. 17 species in three sections (van Gelderen et al. 1994;Grimm et al. 2007;Grimm and Denk 2014).
Likewise, the morphologically distinct Quercus praecastaneifolia strongly resembles a clade of East Asian white oaks (sect. Quercus) that is no longer represented in western Eurasia. It is interesting to note that leaf morphologies similar to extant East Asian species appeared in the Pliocene of Europe but did not persist into modern times. The late Pliocene Q. praecastaneifolia appears to belong to a clade of modern white oaks restricted to the Himalayas to East and Northeast Asia that are relatively frost resistant. In this case, lack of frost and drought tolerance alone are not sufficient to explain why this species went extinct in Europe (cf. Table 2 and table 4 in Svenning 2003). Similarly, Fagus in Frankfurt is virtually identical to modern East Asian species (F. hayatae subsp. pashanica, F. longipetiolata and F. crenata) and has not much in common with the modern western Eurasian Fagus populations. Accepting a late Pliocene (ca. 3.6-2-6 Ma) age for the Pliocene flora of Frankfurt, it is interesting to note that a recent study placed the time of divergence of regional populations of Fagus orientalis from the remainder of F. orientalis and F. sylvatica in the early Pleistocene (1.18-1.87 Ma). Fagus sylvatica s.str. diverged from the remainder of western Eurasian Fagus populations at ca. 817 ka (Gömöry et al. 2018). This is in accordance with other studies (Fagus, Martinetto 2015;Quercus, Bagnoli et al. 2016;Abies, Piotti et al. 2017).